Autoware Diffusion Planner

Overview

The Autoware Diffusion Planner is a trajectory generation module for autonomous vehicles, designed to work within the Autoware ecosystem. It leverages the Diffusion Planner model, as described in the paper “Diffusion-Based Planning for Autonomous Driving with Flexible Guidance” by Zheng et al.

This planner generates smooth, feasible, and safe trajectories by considering:

• Dynamic and static obstacles
• Vehicle kinematics
• User-defined constraints
• Lanelet2 map context
• Traffic signals and speed limits

It is implemented as a ROS 2 component node, making it easy to integrate into Autoware-based stacks. The node is aimed at working within the proposed Autoware new planning framework.

How to use

Currently, some launch files must be changed to run the planning simulator with autoware_diffusion_planner.

(1) /path/to/src/launcher/autoware_launch

diff --git a/autoware_launch/config/control/trajectory_follower/longitudinal/pid.param.yaml b/autoware_launch/config/control/trajectory_follower/longitudinal/pid.param.yaml
--- a/autoware_launch/config/control/trajectory_follower/longitudinal/pid.param.yaml
+++ b/autoware_launch/config/control/trajectory_follower/longitudinal/pid.param.yaml
@@ -6,7 +6,7 @@
     enable_overshoot_emergency: false
     enable_large_tracking_error_emergency: true
     enable_slope_compensation: true
-    enable_keep_stopped_until_steer_convergence: true
+    enable_keep_stopped_until_steer_convergence: false

     # state transition
     drive_state_stop_dist: 0.5
diff --git a/autoware_launch/config/system/diagnostics/planning.yaml b/autoware_launch/config/system/diagnostics/planning.yaml
--- a/autoware_launch/config/system/diagnostics/planning.yaml
+++ b/autoware_launch/config/system/diagnostics/planning.yaml
@@ -11,19 +11,7 @@ units:
           - { type: link, link: /autoware/planning/trajectory_validation }

   - path: /autoware/planning/trajectory_validation
-    type: and
-    list:
-      - { type: link, link: /autoware/planning/trajectory_validation/finite }
-      - { type: link, link: /autoware/planning/trajectory_validation/interval }
-      - { type: link, link: /autoware/planning/trajectory_validation/curvature }
-      - { type: link, link: /autoware/planning/trajectory_validation/angle }
-      - { type: link, link: /autoware/planning/trajectory_validation/lateral_acceleration }
-      - { type: link, link: /autoware/planning/trajectory_validation/acceleration }
-      - { type: link, link: /autoware/planning/trajectory_validation/deceleration }
-      - { type: link, link: /autoware/planning/trajectory_validation/steering }
-      - { type: link, link: /autoware/planning/trajectory_validation/steering_rate }
-      - { type: link, link: /autoware/planning/trajectory_validation/velocity_deviation }
-      - { type: link, link: /autoware/planning/trajectory_validation/trajectory_shift }
+    type: ok

   - path: /autoware/planning/routing/state
     type: diag

(2) /path/to/autoware/src/universe/autoware_universe

```diff diff –git a/launch/tier4_planning_launch/launch/planning.launch.xml b/launch/tier4_planning_launch/launch/planning.launch.xml — a/launch/tier4_planning_launch/launch/planning.launch.xml +++ b/launch/tier4_planning_launch/launch/planning.launch.xml @@ -40,12 +40,34 @@ </include> </group>

• </include>
• </include>
• </group> +

File truncated at 100 lines see the full file

Changelog for package autoware_diffusion_planner

0.48.0 (2025-11-18)

• Merge remote-tracking branch 'origin/main' into humble

• fix(diffusion_planner): fix poses in postprocess (#11618)

  • Refactored tensor_data
  • Fixed to parse_predictions
  • Added an error handling
  • Removed an unused [&]{.title-ref}
  • Added [reserve]{.title-ref}

  * Fixed typecasting from [int]{.title-ref} to [uint32_t]{.title-ref} ---------

• fix: tf2 uses hpp headers in rolling (and is backported) (#11620)

• fix(diffusion_planner): fix [is_segment_inside]{.title-ref} (#11602) Fixed is_segment_inside

• feat(diffusion_planner): add [stopping_threshold]{.title-ref} (#11541)

  • Added stopping_threshold
  • Applied clang-format
  • Fixed stopping logic
  • Added enable_force_stop
  • Fixed create_ego_trajectory
  • Fixed ego_kinematic_state_
  • Added a comment

  * Fixed the condition ---------

• fix(diffusion_planner): change the default parameters (#11520) Changed the default parameters

• refactor(diffusion_planner): remove [use_route_handler]{.title-ref} (#11518)

  • Removed use_route_handler

  * Applied clang-format ---------

• fix(diffusion_planner): wrong agent index (#11454)

  • fix(diffusion_planner): wrong agent index

  * fix(diffusion_planner): correct method call for neighbor agent data retrieval ---------

• fix(diffusion_planner): fix remaining velocity (#11443) Fixed remaining velocity

• refactor(diffusion_planner): fix ego_history (#11395)

  • Fixed ego_history
  • Added error handling
  • Fixed tests
  • Fixed to use Eigen::Isometry3d(mat).inverse()
  • Fixed the function comment

  * Add detailed comments ---------

• feat(diffusion_planner): add acceleration (#11392)

  • Fixed const dimensions
  • Refactored create_neighbor_trajectories
  • Fixed create_ego_trajectory
  • Removed [to_candidate_trajectories_msg]{.title-ref}
  • Removed get_prediction_matrix
  • Removed create_neighbor_trajectories
  • Fixed create_predicted_objects

  * Added velocity_smoothing_window ---------

• refactor(diffusion_planner): move postprocessing internal functions to cpp (#11390)

  • Moved postprocessing internal functions

  * Removed tests related to internal functions ---------

• refactor(diffusion_planner): refactor lane_segment (#11378)

  • Refactored
  • Fixed using
  • Added [epsilon]{.title-ref}
  • Added const

File truncated at 100 lines see the full file